R: Predicting from Smoothing Spline ANOVA Fits with Non-Gaussian...

predict9.gssanova {gss}

R Documentation

Predicting from Smoothing Spline ANOVA Fits with Non-Gaussian Responses

Description

Evaluate smoothing spline ANOVA fits with non-Gaussian responses at arbitrary points, with results on the response scale.

Usage

## S3 method for class 'gssanova'
predict9(object, newdata, ci=FALSE, level=.95, nu=NULL, ...)

Arguments

`object`	Object of class inheriting from `"gssanova"`.
`newdata`	Data frame or model frame in which to predict.
`ci`	Flag indicating if Bayesian confidence intervals are required. Ignored for `family="polr"`.
`level`	Confidence level. Ignored when `ci=FALSE`.
`nu`	Sizes for `"nbinomial"` fits with known sizes. Ignored otherwise.
`...`	Ignored.

Value

For ci=FALSE, predict9.gssanova returns a vector of the evaluated fit,

For ci=TRUE, predict9.gssanova returns a list of three elements.

`fit`	Vector of evaluated fit on response scale.
`lcl`	Vector of lower confidence limit on response scale.
`ucl`	Vector of upper confidence limit on response scale.

For family="polr", predict9.gssanova returns a matrix of probabilities with each row adding up to 1.

Note

For mixed-effect models through gssanova or gssanova1, the Z matrix is set to 0 if not supplied. To supply the Z matrix, add an element random=I(...) in newdata, where the as-is function I(...) preserves the integrity of the Z matrix in data frame.

Unlike on the link scale, partial sums make no sense on the response scale, so all terms are forced in here.

References

Gu, C. (2013), Smoothing Spline ANOVA Models (2nd Ed). New York: Springer-Verlag.